llvm dev - Mar 2005 - [LLVMdev] Recursive Types using the llvm support library

On Sun, 20 Mar 2005, John Carrino wrote:
> On Wed, Mar 09, 2005 at 04:05:32PM +0300, Vladimir Merzliakov wrote:
>>>> Is assert(!NewSTy->isAbstract()) must pass  after this line?
>>>
>>> In this case, yup.
>>>
>> I create test program and assert failed in it:
>>
>> { \2 *, sbyte * }
>
> How do I decode the \2 in this?  I am creating types through this
> interface and I get quite a mess seen below.  And this is relatively
> simple.  Any ideas?
"\2" is an upreference, which refers to a containing type.  For
example,
this type:

  \1*

Is a pointer to pointer to pointer to pointer....  In "{\2*}", the 
upreference refers to the structure type, not the pointer, with \4, it 
means go 4 types up etc.

In practice, this is leads to a horrible mess, as you've noticed.  As 
such, I'd suggest using Module::addTypeName to give these things names so 
that you have a hope to understand them and so they are more compact :)

-Chris

> 	alloca { { int, uint, sbyte*, \4* (), \4* (\4*), { { int, uint, sbyte*,
\4* (), { { int, uint, sbyte*, \4 (), \4 (\4), \8* (\4), \4 (\4) } }* (\4*), \4*
(\4*), \4* (\4*), { { int, uint, sbyte*, \4 (), \4 (\4), \8* (\4), \4 (\4) } }*
(\4*), \4* (\4*, \4*), \4* (\4*, { { int, uint, sbyte*, \4 (), \4 (\4), \8*
(\4), \4 (\4) } }*, { { int, uint, sbyte*, \4 (), \4 (\4), \8* (\4), \4 (\4) }
}*) } }* (\4*), \4* (\4*) } }		; <{ { int, uint, sbyte*, \4* (), \4* (\4*), {
{ int, uint, sbyte*, \4* (), { { int, uint, sbyte*, \4 (), \4 (\4), \8* (\4), \4
(\4) } }* (\4*), \4* (\4*), \4* (\4*), { { int, uint, sbyte*, \4 (), \4 (\4),
\8* (\4), \4 (\4) } }* (\4*), \4* (\4*, \4*), \4* (\4*, { { int, uint, sbyte*,
\4 (), \4 (\4), \8* (\4), \4 (\4) } }*, { { int, uint, sbyte*, \4 (), \4 (\4),
\8* (\4), \4 (\4) } }*) } }* (\4*), \4* (\4*) } }*>:2 [#uses=0]
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://mail.cs.uiuc.edu/mailman/listinfo/llvmdev
>
-Chris

-- 
http://nondot.org/sabre/
http://llvm.cs.uiuc.edu/

On Sun, Mar 20, 2005 at 07:01:55PM -0600, Chris Lattner
wrote:
> On Sun, 20 Mar 2005, John Carrino wrote:
> 
> >On Wed, Mar 09, 2005 at 04:05:32PM +0300, Vladimir Merzliakov wrote:
> >>>>Is assert(!NewSTy->isAbstract()) must pass  after this
line?
> >>>
> >>>In this case, yup.
> >>>
> >>I create test program and assert failed in it:
> >>
> >>{ \2 *, sbyte * }
> >
> >How do I decode the \2 in this?  I am creating types through this
> >interface and I get quite a mess seen below.  And this is relatively
> >simple.  Any ideas?
> 
> "\2" is an upreference, which refers to a containing type.  For
example,
> this type:
> 
>  \1*
> 
> Is a pointer to pointer to pointer to pointer....  In "{\2*}",
the
> upreference refers to the structure type, not the pointer, with \4, it 
> means go 4 types up etc.
> 
> In practice, this is leads to a horrible mess, as you've noticed.  As 
> such, I'd suggest using Module::addTypeName to give these things names
so
> that you have a hope to understand them and so they are more compact :)
I took your advide and used Module::addTypeName which is a great idea
that I didn't know about. 

Now I am getting a seg fault, here

SymbolTable.cpp:251
  if (T->isAbstract()) {
    cast<DerivedType>(T)->addAbstractTypeUser(this);

I am assuming the patch mentioned earlier on the thread will fix this.
I just wanted to ask to make sure before i went ahead and recompiled.

-John

On Sun, 20 Mar 2005, John Carrino wrote:
>> In practice, this is leads to a horrible mess, as you've noticed. 
As
>> such, I'd suggest using Module::addTypeName to give these things
names so
>> that you have a hope to understand them and so they are more compact :)
>
> I took your advide and used Module::addTypeName which is a great idea
> that I didn't know about.
>
> Now I am getting a seg fault, here
>
> SymbolTable.cpp:251
>  if (T->isAbstract()) {
>    cast<DerivedType>(T)->addAbstractTypeUser(this);
>
> I am assuming the patch mentioned earlier on the thread will fix this.
> I just wanted to ask to make sure before i went ahead and recompiled.
You would have to give me more context to say for sure, but updating to 
current CVS can't hurt.

-Chris

-- 
http://nondot.org/sabre/
http://llvm.cs.uiuc.edu/